Evaluasi Teknik Seleksi Fitur terhadap Performa Model Machine Learning pada Dataset Penyakit Sirosis

Abstract

Cirrhosis is a chronic disease of the liver with high levels of morbidity and mortality, so accurate and efficient early detection efforts are needed. The use of Machine Learning (ML) in the health sector is one solution to help the process of classifying and diagnosing diseases based on patient clinical data. However, using all the features in a dataset without selection can cause a decrease in model performance due to the presence of irrelevant or redundant features. Therefore, this study aims to evaluate the effect of feature selection techniques on the performance of ML models such as SVM, Tree, kNN and Random Forest on the cirrhosis dataset. The stages used in this research include the application of several ML models as classification models, as well as the application of various feature selection techniques such as Information Gain (IG), Gain Ratio (GR), and Gini Decrease (GD) and performance evaluation. Model evaluation is carried out by comparing classification results before and after feature selection using accuracy, precision and recall metrics. The research results show that the application of feature selection can significantly improve the performance of the cirrhosis classification model, while reducing the complexity of the model. In addition, the effectiveness of feature selection techniques varies depending on the ML algorithm used, so there is no one feature selection method that is always superior for all models. This research proves that feature selection is an important stage in ML-based medical data processing and can increase the accuracy and efficiency of cirrhosis diagnosis